The infant mortality rate (IMR) is a fundamental and widely studied measure of a population's health. This project will examine a key measurement issue for many studies that examine the IMR: how does variation in the classification of live births influence the measured IMR and comparisons of IMR across time periods, group, and place? Specifically, a death cannot contribute to the IMR unless it is first categorized as a live birth, and this classification may involve some ambiguity. Moreover, as medical technology improves survival at the margins of viability, doctors may become more aggressive in the management of marginally viable fetuses, thereby increasingly classifying these cases as live births. Using Vital Statistics (VS) cohort linked birth/infant death files, birth files, and fetal eath files, we will pursue two specific aims, the first assessing the importance of live birth endogeneity to IMR trends and the second to IMR disparities. We propose to use state-of-the-art statistical methods and several direct and indirect methods to examine these issues.